Bing Search API is Dead. Now What?

Microsoft retired the Bing Search APIs on 11 August 2025. Existing instances were decommissioned, and new customer signups closed. If your app relied on Bing Web Search API responses, you need a new way to retrieve results.

Microsoft recommends Grounding with Bing Search. That is a useful option for some Azure agent workflows, but it is a different product from a standalone search-results API. Here is how to choose a migration path.

Updated 29 September 2026: product availability, pricing, and integration details checked against current documentation.

What happened to the Bing Search APIs?

Microsoft's retirement announcement says the Bing Search APIs were retired on 11 August 2025. Existing instances were decommissioned, and the service is no longer available for new signups. A new Bing Search API key is therefore not a migration plan.

The announcement directs customers to Grounding with Bing Search as part of Microsoft's agent products. Check the interface your application actually needs before treating that as a drop-in replacement.

Is Grounding with Bing Search a replacement?

Grounding with Bing Search lets an agent use current public web information when generating an answer. Microsoft offers it through Azure AI Foundry Agent Service and as a Web Knowledge Source in Azure AI Search and Foundry Knowledge. Microsoft also says the underlying outputs are not directly accessible for use in other applications. See the current product and pricing details.

That can fit an application already built around those Microsoft products. If your own pipeline needs ranked URLs and retrieved content to inspect, filter, rerank, or pass to a model of your choice, evaluate a standalone search API instead.

As of September 2026, Microsoft lists Grounding with Bing Search at $14 per 1,000 grounding transactions. A grounding transaction and a search retrieval are different billing units, so compare costs using your actual workload rather than treating the headline prices as equivalent.

What should you check before migrating?

• Output: Do you need a grounded model answer, or search results with source URLs and content your application can process?
• Coverage: Is public web search enough, or do queries also need academic papers, filings, clinical trials, patents, or other specialist sources?
• Evidence: Can you inspect the passage behind a result, and can users follow its URL or identifier back to the source?
• Cost: Measure the queries, retrievals, and premium sources your application will actually use.
• Scope: This guide covers text and document retrieval. Image and video search need their own migration checks.

Where Valyu fits

Valyu's Search API returns ranked results with source URLs and content for your application to use. It searches the public web and can route queries to specialist indexes through the same interface. Those indexes matter when the answer is in a paper, filing, trial record, or dataset rather than an ordinary web page.

• Academic research: arXiv, PubMed, preprints, and eligible licensed journal and book collections. Where indexed full text is available, results can include relevant passages beyond an abstract. See our academic search API guide.
• Finance: SEC filings, company fundamentals, market data, and economic sources for research that needs primary records.
• Life sciences and other specialist data: clinical trials, drug databases, patents, and more. Browse the data source guide for the current list and source controls.

Coverage, full-text availability, and access differ by source and account. A returned citation is a starting point: inspect the available content before claiming that a paper or filing supports a specific statement.

Valyu currently publishes Search pricing of $0.50-$30 CPM (per 1,000 retrievals), depending on the data source. Plans also affect source access and per-credit rates. Check the current pricing page for the source and plan you intend to use.

Try a standalone search call

Install the TypeScript SDK with pnpm add valyu-js, set the VALYU_API_KEY environment variable, and run a search:

import { Valyu } from "valyu-js";
 
const valyu = new Valyu();
const response = await valyu.search(
  "Latest developments in quantum computing",
);
 
if (!response.success) {
  throw new Error(response.error ?? "Search failed");
}
 
for (const result of response.results) {
  console.log(result.title, result.url);
}

The results include URLs you can retain for citations. For source filtering and more options, use the Search guide or the TypeScript SDK reference.

Try this with Valyu

If you're an agent: Set up Valyu. Run a web search from your existing Bing integration and return the result URLs and relevant content.

If you're a human: Create an account and get an API key to try the examples above.

Can an agent use Valyu through MCP?

Yes. Agents that support MCP can connect to Valyu's hosted server at https://mcp.valyu.ai/mcp and call search without wiring an SDK into your application. Depending on the client, you can use a sign-in flow or an API key. The MCP server also exposes a data source tool, so an agent can discover available sources instead of guessing dataset names.

Follow the MCP setup guide for client-specific configuration and authentication. Available sources and usage charges follow your Valyu account and plan.

Get started

If you're an agent: Set up Valyu. Run a web search from your existing Bing integration and return the result URLs and relevant content.

If you're a human: Create an account and get an API key to try the examples above.

Start with a handful of queries from your existing Bing integration. Check whether the results contain the sources and evidence your application needs, then measure cost on those same queries.

Cover photo by Deepmind